308 pritchard's wedge photometer. 



Here the probable error of the mean is 3.1 div., but that of a single observa- 

 tion is 13.9 div., or three per cent of the whole. This increased error is due 

 almost wholly to the incessant variations of diathermancy of the air, and, as it 

 will be seen by what has preceded, it is at least twelve times the proper instru- 

 mental error. With the aid of the foregoing observations, the reader will have a 

 clearer apprehension of the trustworthiness of the means used in the following 

 determinations. 



Two kinds of measure are taken. The first is of the transmission of the total 



solar radiation ; and in this a bolometer of 1 mm. aperture is placed directly in the 

 path of the beam at 5 meters from the slit. 



In the second, which is for the purpose of studying the absorption spectrum 

 of the wedge, the beam after passing through the wedge and slit is rendered 

 parallel by a horizontal collimator, l (Fig. 4), of 7.5 meters' focus, whence the rays 

 fall on a large flint prism, p, to form a heat spectrum, which is received by a concave 

 mirror, m, of 1.5 meters' focal length, and focused on the 1 mm. bolometer b, which 

 is now attached to the bolometer spectroscope. The light from the siderostat 

 mirror n is kept accurately adjusted at the centre of the screen s' placed before 

 the lens l. The deviations correspond to the wave-lengths M .4, 0\5, 0".6, M .7, 



and 1 M .0, the last being invisible * Whichever kind of measure we take, it is 

 that we are in effect working through four successive equal increments of the glass 

 from U .3 to 6 M .6, and we may as a matter of convenience call each of these 

 increments unity. 



In the prismatic observations a difficulty arises from the great range. The scale 

 of the galvanometer is divided into 1,000 equal parts, of which the central 500 are 

 alone ordinarily used: but the range here, from the small heat in the violet 



ray (0*.4), passing through the thickest part of the wedge, to the great heat in 

 the red ray (0^.7), passing through the thinnest part of it, is much over 10,000 to 1. 

 Such differences, it will be understood, only present themselves in the prismatic 

 investigation ; but on account of them we have been obliged here to narrow the slit 

 in some cases for the full beam to one half its normal aperture, and to reduce the 

 values thus obtained to what they would have been with the 2 mm. slit usually 

 employed. The values thus obtained are overscored thus : TOO. 



* v = 0.001 mm. = 10,000 of Angstrom's scale. 



